Systems and methods for detecting blight and code violations in images

Patent metadata

Field	Value
Publication number	US-12536632-B2
Application number	US-202117383551-A
Country	US
Kind code	B2
Filing date	Jul 23, 2021
Priority date	Jul 28, 2020
Publication date	Jan 27, 2026
Grant date	Jan 27, 2026

Abstract

A system for blight and code violation detection is provided. The system includes a machine learning model that is trained using photographs of properties that are associated with blight and code violations and photographs that are not associated with blight and code violations. A fleet of vehicles, such as trash trucks, are equipped with cameras to take photographs of properties along their routes. The trained model may generate a score for each photograph that indicates whether or not a property is blighted or has code violations. Those properties with a score that exceeds a threshold may be provided to a reviewer who may verify the finding. If the finding is verified, the reviewer may issue a letter or citation to the owner of the property, and positive feedback may be provided to the model. If the finding is not verified, negative feedback may be provided to the model.

First claim

What is claimed:

1 . A method for detecting blighted properties or determining code violations comprising:
receiving training data by a computing device, wherein the training data comprises a first plurality of images of properties with code violations and a second plurality of images of properties without code violations, wherein a property has a code violation when it does not comply with a particular state or local regulation, wherein the properties consist of homes or buildings; extracting a first set of features from the first plurality of images of properties with code violations by the computing device; extracting a second set of features of the second plurality of images of properties without code violations by the computing device; using machine-learning to generate, from the first and second set of features extracted from the first and second plurality of images, a model by the computing device; receiving a third plurality of images by the computing device, wherein each image of the third plurality of images depicts a property other than the properties associated with the first plurality of images or the properties associated with the second plurality of images, wherein each image of the third plurality of images is associated with GPS coordinates, and each image of the third plurality of images was captured automatically by a municipal vehicle upon detection by the municipal vehicle using a location determination component of the municipal vehicle that the municipal vehicle has advanced to the property depicted in the image; and for each image of the third plurality of images:
using the model, generating a score for the property depicted in the image based on the image;
based on the score, determining whether the property depicted in the image has one or more code violations; and
providing the image to a reviewer by the computing device.

2 . The method of claim 1 , wherein each property is one or more of a house, a sidewalk, a road segment, or a tree.

3 . The method of claim 1 , further comprising, for each image of the third plurality of images:
receiving feedback from the reviewer regarding the determination; and adjusting the model based on the received feedback.

4 . The method of claim 1 , further comprising, for each image of the third plurality of images:
identifying one or more features of the image that are associated with the score; for each identified one or more features, detecting an object corresponding to the one or more features; and generating a report for the property associated with the image, wherein the report identifies the detected object.

5 . The method of claim 4 , further comprising:
for each detected object, determining a portion of the score that is attributable to the detected object; and providing the determined portion for each detected object in the report.

6 . The method of claim 4 , further comprising, for each image of the third plurality of images:
providing the report to an address associated with the property depicted in the image.

7 . The method of claim 6 , wherein the report comprises a fine or a required action.

8 . The method of claim 1 , wherein the municipal vehicle is a garbage truck.

9 . The method of claim 1 , further comprising, weighting the score generated for the property depicted in the image by the scores generated for properties that are neighboring the property depicted in the image to generate a relative score, and determining whether the property depicted in the image has one or more code violations based on the relative score.

10 . The method of claim 1 , further comprising, based on the one or more code violations, recommending one or more repairs to the property depicted in the image.

11 . A system for detecting blighted properties or code violations comprising:
at least one processor; and a memory storing instructions that when executed by the at least one processor cause the at least one processor to:
receive training data, wherein the training data comprises a first plurality of images of properties with code violations and a second plurality of images of properties without code violations, wherein a property has a code violation when it does not comply with a particular state or local regulation, wherein the properties consist of homes or buildings;
extract a first set of features from the first plurality of images of properties with code violations;
extract a second set of features of the second plurality of images of properties without code violations;
use machine-learning to generate, from the first and second set of features extracted from the first and second plurality of images, a model;
receive a third plurality of images, wherein each image of the third plurality of images depicts a property other than the properties associated with the first plurality of images or the properties associated with the second plurality of images, wherein each image of the third plurality of images is associated with GPS coordinates, and each image of the third plurality of images was captured automatically by a municipal vehicle upon detection by the municipal vehicle using a location determination component of the municipal vehicle that the municipal vehicle has advanced to the property depicted in the image; and
for each image of the third plurality of images:
using the model, generate a score for the property depicted in the image based on the image;
based on the score, determine whether the property depicted in the image has one or more code violations; and
provide the image and the determination to a reviewer.

12 . The system of claim 11 , wherein each property is one or more of a house, a sidewalk, a road segment, or a tree.

13 . The system of claim 11 , further comprising, for each image of the third plurality of images:
receiving feedback from the reviewer regarding the determination; and adjusting the model based on the received feedback.

14 . The system of claim 11 , further comprising, for each image of the third plurality of images:
identifying one or more features of the image that are associated with the score; for each identified one or more features, detecting an object corresponding to the one or more features; and generating a report for the property associated with the image, wherein the report identifies the detected object.

15 . The system of claim 14 , further comprising:
for each detected object, determining a portion of the score that is attributable to the detected object; and providing the determined portion for each detected object in the report.

16 . The system of claim 14 , further comprising, for each image of the third plurality of images:
providing the report to an address associated with the property depicted in the image.

17 . The system of claim 16 , wherein the report comprises a fine or a required action.

18 . The system of claim 11 , wherein the municipal vehicle is a garbage truck.

19 . A method for automatically detecting blighted properties or determining code violations comprising:
attaching a camera and a location determination component to a municipal vehicle associated with a route; for each property of a plurality of properties along the route, wherein the properties of the plurality of properties consist of homes or buildings:
detecting that the municipal vehicle has moved to a next property of the plurality of properties on the route by a computing device using the location determination component; and
in response to the detection

Assignees

• Univ Alabama

Inventors

• Johnson Erik Barry
• Moore Brendan August

Classifications

• G06V20/58

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06T2207/30188

  Vegetation; Agriculture · CPC title

• G06T2207/30184

  Infrastructure · CPC title

• G06T2207/20081

  Training; Learning · CPC title